Search

Kids use the computer assisted language learning systems to learn English. The contents of the system are well designed and kids enjoy them. From Cognitive Psychology we found gestures played useful role in learning so we developed the language learning system utilizing gestures. The system provides similar contents as the existing system but tries to enforce users to follow gestures related to given words. We compared the proposed system with an existing one in terms of memorizing test scores. The average improvement achieved using the proposed system was little better than one achieved using the existing system.

One of the major issues in mixed reality multi-agent systems is synchronization of display, which can adversely affect system performance and hinder user interaction. Real time response from the system cannot be achieved because of the aforementioned issues. If the content displayed on agents is complex and cannot be feasibly rendered on a single agent, then a better strategy is to divide the contents among multiple agents. In this way, only a fraction of the entire contents is rendered by each agent. In this paper, two alternative techniques for multi-agent based content management are proposed, namely, full contents on agents (FCOA) and partial contents on agents (PCOA). In FCOA, each agent in the multi-agent system renders all the contents and only a specific part of the contents is displayed depending upon the usage scenario. In PCOA the agents receive partial contents from the mediator. A comparative study has been presented in this paper to identify the pros and cons of each method.

Advertisements and campaigns on multiscreen with digital monitors are recently becoming very common in public places. However current virtual reality technology has limits in the feeling of physical space which is an important factor for feeling of real space. We present a study on adaptive audio reproduction for physical and spatial immersion with mixed reality technology. Especially we tried to adjust the distortion of audio signal spectrum that can occur at users dynamic locations. As a result, we need the differentiated adjustment audio signal in frequency as well as loudness according to a distance between a user and a speaker.

In this paper, we present image genre classification system, called IGC. The proposed system categorizes images into one of three genres, such as art, photo, or cartoon images. The images features are extracted using standard MPEG-7 visual descriptors, after which they are trained using Neural Networks. The simulation results show that the proposed system successfully classifies images into correct classes with the rate of over 85% depending on the employed features.

Today cartoon images take more portion of digital multimedia than ever as we notice this phenomenon in the entertainment business. With the explosive proliferation of cartoon image contents on the Internet, we seem to need a classification system to categorize these cartoon images. This paper presents a new approach of cartoon image classification based on cartoonists. The proposed cartoon image classification system employs effective MPEG-7 descriptors as image feature values and learns features of particular cartoon images, and classifies the images as multiple classes according to each cartoonist. In the performance simulation we evaluate the effectiveness of the proposed system on a large set of cartoon images and the system successfully classifies images into multiple classes with the rate of over 90%.

This paper presents an overview of a distributed data mining system developed according to two approaches; 1) distributed data modeling and 2) distributed decision making. A communication mechanism between multiagents was established in the system, in order to achieve two approaches to deal with information processing in a distributed way. An algorithm for each approach has been modified from the original forms of decision tree building/ making algorithms so that it can work under multi-agents computing environments as well as single computing environments. This system is evaluated over a test bed of US census bureau statistical data set collected in the 1990s. For a comparison of the distributed analysis version with the single version with respect to analysis results, the data set was disjoined into two-four different databases which were deployed in different local sites.

In this paper, we propose a robust method for the process of localization of a mobile robotthrough a vision system. The mobile robot is a compact system consisting of an embedded board and a fish-eye camera. The fish-eye camera looks upward to capture ceiling images. The camera provides a sequence of images for the process of detection and tracking ceiling features. These features are used like natural landmarks to detect the state of translation and rotation of the robot. Our method requires less computational power and resources for a robot, and thus can be used at home. The results produced in this study showed the advantages of our method in terms of both speed and accuracy.

The paper presents a method for content based image retrieval (CBIR) using an adaptive image classification with Radial Basis Function networks. It supports geographical image retrieval over digitized historical aerial photographs, in a digital library, which are gray-scaled and low-resolution images. CBIR is achieved on the basis of texture feature extraction and image classification. Feature extraction methods for geographical image analysis are Gabor spectral filtering and Laws' energy filtering, which are the most widely used in image classification and segmentation. Image classification supports effective CBIR through composite classifier models dealing with multi-modal feature distribution. The method is evaluated over a digital library that contains collections of thousands of small-sized texture tiles obtained from large-sized aerial photograph images with geographical features.

This paper describes the implementation of a distributed data mining system. The system consists of a web server, a pre-processor for data preparation, a mediator, and agents. A distributed learning algorithm of a decision tree in an agent-mediator communication mechanism is the most important and difficult to achieve the distributed data mining in this system, in the view of implementation. The algorithm has successfully been implemented with several techniques. Its implementation is presented in a UML (Unified Modeling Language) sequence diagram.

This paper presents an approach for target movement prediction by using Genetic Algorithms to generate the population of movement generation operators In this approach, we use objective functions, not derivatives or other auxiliary knowledge, and apply probabilistic transition rules, not deterministic rules, for target movement prediction Its performance has been experimentally evaluated through several experiments.

In our knowledge-based society, collaboration environments define a key Knowledge Management trend. The critical step in the implementation of this step is to capture the collaboration surrounding visual reports (e.g., charts) used in decision-making. Indeed, the ability to provide an accurate and integrated visualization of the data spaces in order to support organizational decision-making is a crucial component of modern information-rich enterprises. This ability can be significantly improved through the introduction of collaborative decision support tools tailored for visualization. This paper presents the development of a class of such tools, called DIVI (Distributed Intelligent Visual Interfaces). The DIVI approach is based on the application of distributed visualization nodes and a mechanism for visual information abstraction/generalization. DIVI technology is envisioned to have a major impact on those applications requiring information to be collected from different data sources for collaborative group decision-making. Typical applications include planning systems, logistics, computer-aided medical group diagnosis, and educational services.

A new background modeling approach is presented in this paper. In most background modeling approaches, input images are categorized into foreground and background regions using pixel-based operations. Because pixels on the input image are considered individually, parts of foreground regions are frequently turned into the background, and these errors cause incorrect foreground detections. The proposed approach reduces these errors and improves the accuracy of a background modeling. Each input image is categorized into three regions in the proposed approach instead of two regions, background and foreground regions. The proposed approach divides traditional foreground regions into two sub-regions, intermediate background and foreground regions, using activity measurements computed from optical flows at each pixel. The other difference of the proposed approach is grouping pixels into objects and using those objects at the background updating procedure. Pixels on each object are turned into the background at the same rate. The rate of each object is computed differently depending on its category. By controlling the rate of turning input pixels into the background accurately, the proposed approach can model the background accurately.

This paper presents a novel method for estimating the coordinates of a 3D object using the four vertices of a quadrangle and the camera motion parameters. Estimation of 3D object coordinates from 2D images of video is a studied problem in augmented reality. However, most solutions are dependent on fiducial markers in video or known coordinate systems which are required with superimposition of virtual object on frames. In this paper, we begin with the fact that the rectangular objects in 3D real world are projected the perspective quadrangle onto image planes. We can estimate 3D object coordinates from 4 vertices of quadrangular objects through transformation of image coordinates. The camera motion parameters between pairs of successive frames in a sequence are calculated using epipolar geometry.

This paper presents a distributed approach to build decision trees in a lock step manner with each node proposing an attribute on which to split A central mediator chooses the attribute, among the candidates, with the highest information gain The chosen split is then effectively communicated to the other agents to partition their data The distributed decision tree approach is performed on the agent based architecture dealing with distributed databases This paper mainly focuses on the evaluation of the system performance in distributed data mining Even though there are several trials suggesting algorithms of distributed data mining, few efforts have made on the definition of the system performance It is very important to define the performance for the further development of distributed data mining.

Radial basis function (RBF) networks have been successfully applied to function interpolation and classification problems among others. In this paper, we propose a basis function optimization method using a mixture density model. We generalize the Gaussian radial basis functions to arbitrary covariance matrices, in order to fully utilize the Gaussian probability density function. We also try to achieve a parsimonious network topology by using a systematic procedure. According to experimental results, the proposed method achieved fairly comparable performance with smaller number of hidden layer nodes to the conventional approach in terms of correct classification rates.

The MPEG-7 texture description tools are applied to the retrieval of the historical aerial photographs. These aerial photographs had been digitized into gray-scaled and low-resolution images to build a digital library. The aerial image retrieval over the digital library is achieved on the basis of 1) texture feature extraction methods, which are homogeneous texture, texture browsing and edge histogram in the MPEG-7 texture descriptors, and 2) their corresponding similarity measurements. The image retrieval methods are evaluated over a collection of historical aerial photographs. The paper presents the retrieval results of these texture feature extraction methods.

Digital multimedia data is dramatically being increased everyday since the Internet became popular. This increment in multimedia data increases adult image contents to the Internet as well. Consequently, a large number of children are exposed to these X-rated contents. In this paper, we propose an efficient classification system that can categorize input images into adult or non-adult images. The simulation shows that this system achieved 95% of the true rate whereas it reduces the false positive rate below 3%.

This paper focuses on the development of an interactive 3D visualization environment for displaying predicted movements by genetic evolution. The integration of interactive 3D visualization and the genetic search approach can relieve the burden of image exploitation tasks occurring due to an explosion of imagery data obtained in battle fields. The visualization component was implemented using In3D – a cross-platform Java class library designed to enable the creation of interactive 3D visualizations of complex information spaces.

The paper presents a novel image (feature) interpolation method to reinforce the adaptive object recognition system The system deals with texture images in a sequence according to changing perceptual conditions When it recognizes several classes of objects under variable conditions, a fundamental problem is that two or more classes are overlapped on the feature space This interpolation method is useful to resolve the overlapping issue.

This paper presents a prediction method for an efficient on-line model modification in adaptive texture image recognition issue. The approach builds a close-loop interaction between object recognition and model modification systems. Object recognition applies an advanced RBF classifier in order to recognize objects on a current image of a sequence. Model modification manipulates the RBF classifier models by changing the structure of the classifier or/and parameters of classifier nodes in order to adapt to changing situations. For efficient model modification, this work includes modification through prediction that classifier models are changed in advance for adaptation if it is possible to predict the modification patterns. For experimentation, the change of texture characteristics has been investigated over a sequence of texture images acquired under dynamic perceptual conditions. Texture characteristics on images in a sequence have been extracted by Gabor spectral filtering, Laws’ energy filtering and Wavelet Transformation filtering. The results of the investigation justify the need for an on-line model modification over the entire sequence of images in order to preserve the system recognition capability, and present the possibility of prediction by finding partial patterns of texture characteristics change. According to experimental results, it is seen that the prediction in model modification can enhance the competence of the system through experimentations.

The paper presents a content based image retrieval approach with adaptive and intelligent image classification through on-line model modification It supports geographical image retrieval over digitized historical aerial photographs in a digital library Since the historical aerial photographs are grayscaled and low-resolution images, image retrieval is achieved on the basis of texture feature extraction Feature extraction methods for geographical image retrieval are Gabor spectral filtering, Laws' energy filtering, and Wavelet transformation, which are all the most widely used in image classification and segmentation Adaptive image classification supports effective content based image retrieval through composite classifier models dealing with multi-modal feature distribution The image retrieval methods presented in the paper are evaluated over a test bed of 184 aerial photographs The experimental results also show the performance of different feature extraction methods for each image retrieval method.

This paper introduces a novel paradigm of privacy preserving mining for distributed databases. The paradigm includes an agent-based approach for distributed learning of a decision tree to fully analyze data located at several distributed sites without revealing any information at each site. The distributed decision tree approach has been developed from the well-known decision tree algorithm, for the distributed and privacy preserving data mining process. It is performed on the agent based architecture dealing with distributed databases in a collaborative fashion. This approach is very useful to be applied to a variety of domains which require information security and privacy during data mining process.

We propose a new type of preconditioners for symmetric Toeplitz system Tx = b . When applying iterative methods to solve linear system with matrix T, we often use some preconditioner C by the preconditioned conjugate gradient (PCG) method [3]. If T is a symmetric positive definite Toeplitz matrix, two kinds of preconditioners are investigated: the “optimal” one , which minimizes ||C – T ||F, and the “superoptimal” one , which minimize ||I – C−−1T ||F [8]. In this paper, we present a general approach to the design of Toeplitz preconditioners based on the optimal investigating and also preconditioners C with preserving the characteristic of the given matrix T. Fast all resulting preconditioners can be inverted via fast transform algorithms with O(N log N) operations. For a wide class of problems, PCG method converges in a finite number of iterations independent of N so that the computational complexity for solving these Toeplitz systems is O(N log N ) [2].

This paper presents texture feature extraction and selection methods for on-line pattern classification evaluation. Feature selection for texture analysis plays a vital role in the field of image recognition. Despite many approaches done previously, this research is entirely different from them since it comes from the fundamental ideas of feature selection for image retrieval. The proposed approach is capable of selecting the best features without recourse to classification and segmentation procedures. In this approach, probability density function estimation and a modified Bhattacharyya distance method are applied for clustering texture features of images in sequences and for comparing multi-distributed clusters with one another, respectively.

This paper presents preliminary works on an agent-based approach for distributed learning of decision trees. The distributed decision tree approach is applied to intrusion detection domain, the interest of which is recently increasing. In the approach, a network profile is built by applying a distributed data analysis method for the collection of data from distributed hosts. The method integrates inductive generalization and agent-based computing, so that classification rules are learned via tree induction from distributed data to be used as intrusion profiles. Agents, in a collaborative fashion, generate partial trees and communicate the temporary results among them in the form of indices to the data records. Experimental results are presented for military network domain data used for the network intrusion detection in KDD cup 1999. Several experimental results show that the performance of distributed version of decision tree is much better than that of non-distributed version with data collected manually from distributed hosts.

This paper discusses a visualization technique integrated with inductive generalization. The technique represents classification rules inferred from data, as landscapes of graphical objects in a 3D visualization space, which can provide valuable insights into knowledge discovery and model-building processes. Such visual organization of classification rules can contribute to additional human insights into classification models that are hard to attain using traditional displays. It also includes navigational locomotion and high interactivity to facilitate the interpretation and comparison of results obtained in various classification scenarios. This is especially apparent for large rule sets where browsing through textual syntax of thousands of rules is beyond human comprehension. Visualization of both knowledge and data aids in assessing data quality and provides the capability for data cleansing.

This paper presents an agent-based distributed data mining approach dealing with heterogeneous databases located at different sites. It introduces a modified decision tree algorithm on an agent based framework, which produces an accurate global model without transferring data between agents. The novel approach is evaluated over a test bed of texture feature data of 184 aerial photograph images. The experimental results show that the distributed version with more agents outperforms the version with fewer agents when the rule generation from the large database is not complicated.

The paper focuses on the adaptive segmentation of aerial images for aerial surveillance. The adaptive segmentation is achieved by the cooperation of on-line model modification and model based image segmentation through RBF neural network classifier. The on-line model modification allows the RBF classifier to adapt to the changes of geographical features on the aerial images. In addition, the Gabor filtering method for feature extraction is proposed in this experiment to discriminate between geographical features for better image segmentation.

This paper is focused on the development of an adaptive NN-RBF classifier for the recognition of objects. The classifier deals with the problem of continuously changing object characteristics used by the recognition processes. This characteristics change is due to the dynamics of scene-viewer relationship such as resolution and lighting. The approach applies a modified Radial-Basis Function paradigm for model-based object modeling and recognition. On-line adaptation of these models is developed and works in a closed loop with the object recognition system to perceive discrepancies between object models and varying object characteristics. The on-line adaptation employs four model evolution behaviors in adapting the classifier's structure and parameters. The paper also proposes that the models modified through the on-line adaptation be analyzed by an off-line model evaluation method (Gaussian mixture density estimates).